As we approach another round of grand fuzziness in predictions of all sorts, e.g. for president’s elections in some corner of the world, it would be wise to not only take all (and I mean all) of Superforecasting to heart but also to consider helping extending the science of the trade.
By helping me out in finding pointers and content on, and subsequently developing on, the use of fuzzy logic in predictions. As ‘current’ truth values of future states of the world are all quite possible, and going forward even mutually exclusive states may, e.g. on some news, all become more likely, with combined likelihoods rising over 100%. Where FL can play a role to keep track. And we may have to revisit (practical use of) Markov chains with suitable noise-around-parameters built in… But let’s focus on FL first.
Of course, when the End Date, the horizon for some prediction timeslot nears, the choices will be driven to 100/0 — where the crazy idea of random selection (of ‘balls from a pot’) with replacement … with double replacement … [even tinkered with the idea of replacing the non-drawn colour with the drawn one every pick; was hard to think through] may come closer to the idea of starting with some hardly-educated guess and nudging either way on all news points as one goes along; doing a (much-)sort-of random walk from 50/50 to 100/0.
So, if you’d have info on the viability of either approaches, please do drop a note…! Already:
[Free city map dispenser; Delft]